The title of this post is polemical, so I’ll start by telling you that I like using generative AI. I have a ChatGPT subscription. I use it all the time for all kinds of things. Some of them are Letters Blogatory related. ChatGPT has helped me design my new search functionality, and it helped me improve my blog email spam filtering. A lot of my uses aren’t law- or blog-related. When I have a question about French usage that Duolingo won’t answer (“what’s the difference between avoir envie de and vouloir?“) ChatGPT has useful things to say. When I see a new kind of mushroom in the back yard, ChatGPT can tell me all about it. It’s as useful as Google was in its heyday.

But I think generative AI should have only a very limited role in the legal writing process, and then only at the end, after you’ve gotten your ideas on paper (well, you know what I mean) in a form you like.

Reason #1. In principle and still to a large extent in practice, the practice of law is oral, not written. Civil jury trials, and even civil bench trials, are rarer than ever. But a lot of our work gets done in oral hearings on motions for summary judgment, or motions to dismiss for failure to state a claim, or jurisdictional motions. When you’re arguing one of those motions, you have to know your case inside and out. You have to know what the cases say. You have to know the facts of the cases and be ready to engage in some back-and-forth with the judge about hypotheticals and distinctions. You have to think on your feet. Writing, and rewriting, and rewriting again is the best way I know to really learn the facts of your case and the landscape of prior cases that are going to be relevant. If you delegate too much of the drafting process to your AI helper, or if you delegate too much of the process of summarizing and synthesizing the material, I do not think you will have done the work that allows you to be really effective at the oral hearing.

Reason #2. This is related to reason #1. Sometimes the relationship between writer and reader is counsel and judge. Sometimes the relationship between writer and reader is blogger and surfer. Sometime the relationship is student and teacher. But there is always a relationship, and even when we don’t know who our readers are, we should write as if there were a real relationship. I’m not going to be annoying and turn this post into a post on rhetoric—ethos just as important to persuasion as pathos and logos.¹Herbert Stern called ethos your “number one” in his great book, Trying Cases to Win. I don’t have the book to hand and I may be misremembering just what Stern called it. The example I remember from Stern’s book is the prosecutor who addresses the court with the jury watching and says his name is so-and-so and that he represents the United States of America. That’s pretty impressive. It carries with it some built-in oomph. Then there’s the defense lawyer, who, also in the presence of the jury when the court is ready to take a recess, walks over to the prosecutor, puts his hand on his elbow, and says, “now Jack, about those documents …” The prosecutor is just another lawyer after all, trying to put one over on the jury just like his buddy the defense lawyer. Oh no! the prosecutor thought. He’s stealing my number one! I question whether an argument written or mostly written by an AI helper can have the authenticity that an argument written by you can have. That’s partly because legal arguments are not just about “the rules” but about the reasons for the rules, which often have to do with fairness, equity, and justice. I care a lot about what my fellow humans think is fair or just but not at all about what ChatGPT predicts a human writer would write about fairness or justice. You may say, “what difference does it make, since the judge won’t be able to tell whether I wrote the argument or an AI helper wrote it.” My response: if you’re right, have you really done everything you can to develop your own voice as a writer?

Reason #3. Writing is one of the best ways to express one’s self. Don’t misunderstand me. Some may think that people have mature and well thought-out ideas in their minds or their hearts that are waiting to be expressed. I think instead that people have a lot of mushy and poorly thought-out ideas in their minds and their hearts that can become mature and well though-out ideas if we have the discipline to write about them, then revise what we wrote, then revise it again. Writing is so integral to thinking and to personal and professional growth and satisfaction that the prospect of outsourcing it strikes me as very bad news for our profession and for ourselves as thinking and feeling people.

Reason #4. We lawyers are always promoting ourselves. A satirist might portray a lawyer as a kind of peacock. There are lots of ways to promote oneself, from tacky billboards to entries in “Super Lawyers” to giving talks to writing law review articles. But if one of the ways you promote yourself is by writing, and if you’re not really writing the things you are publishing under your name, you might ask yourself whether you’re satisfied that you’re being honest with your readers.

Just for fun, I decided, after writing this post, to ask ChatGPT to write one using the basic ideas from what I had written. Here was my question:

Write a blog post that would be suitable for publication on Letters Blogatory arguing that the use of generative AI in legal blogging is a bad idea, because it does not help the writer become a better writer; because it does not help the writer learn to think; and because it is inauthentic and thus cannot really persuade.

And here is what I got. (My prompt gave the bot a little bit of a head start, because I fed it the conclusion I wanted and the main reasons in summary).

Title: Why Generative AI Is A Bad Idea For Legal Blogging

In the last year, legal blogging—like so much of our professional world—has been touched by the rise of generative AI. The promise is a tempting one: rapid production of readable content, the automation of routine tasks, and more time for “thought leadership” freed from the drudgery of drafting. But on closer inspection, using generative AI to write our legal blog posts is not just an innocuous timesaver. It actively undermines three core virtues of legal writing: the writer’s craft, the writer’s capacity for critical thinking, and the authenticity that makes for genuine persuasion.

1. It Does Not Improve The Writer’s Craft

No one becomes a skilled writer by outsourcing the essence of writing to an algorithm. Good writing is an intellectual exercise that demands attention to structure, style, and voice. It’s a process of constant iteration and self-critique: test a phrase here, find a better synonym there, reorganize a paragraph to lead the reader more naturally toward the conclusion. This painstaking (and yes, sometimes painstakingly slow) process is where the true growth happens. By letting a machine generate the first draft, we cut ourselves out of that creative loop. We reduce ourselves to lightly editing a machine’s work rather than developing our own stylistic flair. Over time, the muscles of composition and nuance atrophy, and what we lose is more than idiosyncratic style—we lose the personal command over language that marks an expert legal writer.

2. It Stunts The Writer’s Capacity For Critical Thinking

Much of the value in legal blogging comes not from the final text, but from the thinking that goes into it. To draft a good post, you have to read the sources carefully, weigh arguments, test hypothetical counterexamples, and consider how best to explain a nuanced legal point. That intellectual deliberation is partly internal and partly external: you learn to think by struggling to find the right words. Delegating the creation of your written words to a machine is like skipping over a crucial phase of mental labor. You’re never forced to wrestle with complexity in the same way; you’re never forced to ask yourself whether a particular concept should be framed in a different light. Instead, you accept the machine’s neatly packaged (but superficial) phrasing. Without that reflective struggle, the true essence of legal thought—the ability to reason clearly, to anticipate counterarguments, to distinguish and analogize cases, to dig into difficult or unfamiliar concepts—is never fully cultivated. The result? A writer who may have more content to publish, but who has not actually grown more thoughtful or analytically adept.

3. It Is Inauthentic And Thus Cannot Truly Persuade

Perhaps the most important reason to reject generative AI in legal blogging is also the simplest: effective legal persuasion is, at its core, an act of authentic communication from one human to another. Our readers—whether they be clients, other lawyers, judges, or interested laypersons—are not merely scanning black-letter words for information; they’re judging the credibility, trustworthiness, and sincerity of the person behind the blog. When the words themselves are the product of an indifferent algorithm, what assurance does the reader have that the argument is genuinely believed and carefully considered? The voice becomes hollow. The reasoning, even if correct in some mechanical sense, is not rooted in a real person’s considered judgment.

Authenticity matters. The unique voice and perspective of a human author carry weight. They let the reader feel they are engaging in a conversation, not simply reading a bland recitation. Authenticity invites trust. If you hand that over to a machine, you forfeit an essential ingredient of persuasion. Your words might pass the spell-check and sound “professional,” but will they convince a skeptical mind? True persuasion emerges from a writer who owns their arguments, who has grappled with complexity and chosen their stance. A machine can crank out sentences, but it cannot imbue them with conviction. And it is conviction—earnest, informed conviction—that ultimately moves readers and shapes debate in the legal world.

In Conclusion

Generative AI may have its place in certain areas of legal practice—research assistance, automated contract drafting, or data analysis—but using it to write our blogs? That’s a step too far. When we rely on a machine to craft our public-facing commentary, we give up the very things that make such commentary worthwhile: our development as better writers, our growth as sharper thinkers, and the authenticity that gives our arguments persuasive force.

Legal blogging should remain a human pursuit. It should remain a space where we hone our language and reasoning skills, wrestle with difficult ideas, and speak in our own voices. By declining to let a machine write our blogs, we preserve not just a mode of expression but also a critical dimension of professional and intellectual development in the legal communihty.

Look, this is clearer, more cohesive, and generally better than a lot of adult humans and a lot of lawyers could do. That’s a problem. When I read it, it screams out “bot!” to me, but that might be because I know it’s written by a bot and because I know what I sound like. It’s astonishing and somehow magical that a digital computer can do this. I’m reminded of something that occurred to me while I was writing about the “elephant habeas” cases. The lawyers claiming to represent the elephants claimed that elephants satisfied the criteria of moral personhood. I asked: “Now, if we are doing philosophy, I could turn that around and ask whether human beings, even competent adult human beings, always satisfy the usual criteria, either in terms of moral agency or in terms of personal identity.” One response to the competence of ChatGPT’s output and the mediocrity of lots of real human writing is to ask whether machines are as intelligent as people; another is to ask whether people are as intelligent as we take ourselves to be.

• 1
  Herbert Stern called ethos your “number one” in his great book, Trying Cases to Win. I don’t have the book to hand and I may be misremembering just what Stern called it. The example I remember from Stern’s book is the prosecutor who addresses the court with the jury watching and says his name is so-and-so and that he represents the United States of America. That’s pretty impressive. It carries with it some built-in oomph. Then there’s the defense lawyer, who, also in the presence of the jury when the court is ready to take a recess, walks over to the prosecutor, puts his hand on his elbow, and says, “now Jack, about those documents …” The prosecutor is just another lawyer after all, trying to put one over on the jury just like his buddy the defense lawyer. Oh no! the prosecutor thought. He’s stealing my number one!